ICPL: Intelligent Cooperative Planning and Learning for Multi-agent Systems
Abstract
The research objective was to develop a new planning approach for teams of multiple UAVs that tightly integrates learning and cooperative control algorithms at multiple levels of the planning architecture. The research results enabled a team of mobile agents to learn to adapt and react to uncertainty in situational awareness and unforeseen future events and thus successfully complete their missions in geographically extended and uncertain theaters of operation. Among application areas, we considered learning approaches for persistent patrolling games, in which one class of agents places point targets in a given region, and a second class of agents seeks to minimize the time necessary to discover such targets, using limited-range sensors. We analyzed equilibrium strategies in this class of problems and their stability. Our e orts provide a fundamental theory and architecture for the design of intelligent cooperative control systems for heterogeneous teams and the demonstration of the value of the theory through software and hardware experiments.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 29, 2012
- Accession Number
- ADA565746
Entities
People
- Emilio Frazzoli
- Jonathan How
- Nicholas Roy
Organizations
- Massachusetts Institute of Technology